The present invention relates generally to the field of data processing, and more particularly to processing un-typed string literal data in triples.
Data type information can improve the quality of data processing results. For example, data type information provides useful information when generating candidate answers in a question answering system. Unfortunately, when information appears in string literal form (e.g., in a database) data type information is generally unavailable.
Triple stores, where data is stored as a flat table of “triples” that each comprise a subject, predicate and object, offer a flexible alternative to relational databases. The subject and object are often referred to as “entities” and the predicate defines the (typically one-way) relationship between those entities. For example, the triples “< . . . /resource/France>< . . . /property/city>< . . . /resource/Paris>” and “< . . . /resource/France>< . . . /property/city>< . . . /resource/Lyon>” indicate that the entities ‘Paris’ and ‘Lyon’ are cities of the entity ‘France’. Triple stores provide increased flexibility and do not require the generation of schemas and linking tables previous to conducting information processing tasks on the stored data.